Information processing device, information processing method, and computer program product for evaluating performance of machine learning model

An information processing method implemented by a computer, the method comprising:

• generating first data to be input to one or more models, each of the one or more models receiving input of traffic flow data and deducing an effect of a case where a measure is executed, the traffic flow data representing a traffic flow of plural roads on which a vehicle runs, the traffic flow data varying depending on whether the measure is executed or not, the first data representing the traffic flow data in a case where the measure is not executed;
• receiving input of one or more first parameters affecting the effect, each first parameter being a parameter used in a data estimation process of estimating second data representing the traffic flow data in a case where the measure is executed;
• executing the data estimation process by using the first parameter received by the receiving of input, and generating the second data;
• estimating the effect by performing an effect estimation process of comparing the first data and the second data, the effect estimation process being different from a process of deducing the effect using the one or more models;
• learning the one or more models by using learning data including the first data and the second data;
• evaluating performance of the one or more models learned by the learning, by comparing a first effect and a second effect, the first effect being the effect deduced by the learned one or more models, the second effect being the effect estimated by performing the effect estimation process;
• select one of the learned one or more models having higher performance than other models;
• apply the selected machine learning model to traffic flow data obtained when the measure is executed, to deduce the effect of the measure;
• and
• output, via a display device, the effect of the measure deduced by the selected machine learning model.

A computer program product comprising a non-transitory computer-readable recording medium on which a program executable by a computer is recorded, the program instructing the computer to:

• generate first data to be input to one or more models, each of the one or more models receiving input of traffic flow data and deducing an effect of a case where a measure is executed, the traffic flow data representing a traffic flow of plural roads on which a vehicle runs, the traffic flow data varying depending on whether the measure is executed or not, the first data representing the traffic flow data in a case where the measure is not executed;
• receive input of one or more first parameters affecting the effect, each first parameter being a parameter used in a data estimation process of estimating second data representing the traffic flow data in a case where the measure is executed;
• execute the data estimation process by using the received first parameter and generate the second data;
• estimate the effect by performing an effect estimation process of comparing the first data and the second data, the effect estimation process being different from a process of deducing the effect using the one or more models;
• learn the one or more models by using learning data including the first data and the second data;
• and
• evaluate performance of the learned one or more models, by comparing a first effect and a second effect, the first effect being the effect deduced by the learned one or more models, the second effect being the effect estimated by performing the effect estimation process;
• select one of the learned one or more models having higher performance than other models;
• apply the selected machine learning model to traffic flow data obtained when the measure is executed, to deduce the effect of the measure;
• and
• output, via a display device, the effect of the measure deduced by the selected machine learning model.